Using CLM(ED) as a basis for representing carbon cycling dynamics in tropical forests

Transcription

1 Using CLM(ED) as a basis for representing carbon cycling dynamics in tropical forests Jennifer A. Holm & Ryan G. Knox Charles D. Koven, William J. Riley, Rosie A. Fisher, Stefan Muszala, Jeffrey Q. Chambers, Niro Higuchi March 2 nd 2015, CESM LMWG Meeting

2 Old-growth tropical forests are responsible for a large portion of the terrestrial carbon sink. In order to improve the quantification of the carbon cycle there is a strong emphasis on incorporating improved vegetation structure and competition in land-surface modeling. Uncertainties in land carbon flux from tropical regions Climate change pushing tropical ecosystems into novel regimes Historical Future Friedlingstein et al. 2014, Nature Geoscience, (CMIP5 analysis) Figure courtesy of Charlie Koven

3 What are the pros to demographic and dynamic vegetation modeling? - Includes disturbances, bioclimatic constraints - Competition for light, water, nutrients, climate dependent - Dynamic mortality and establishment CLM prior to Ecosystem Demography Model (ED) CLM4.5 BGC (lack of demographics) Vegetation treated as a big leaf model Objectives: Testing, diagnosing, and fine-tuning the newly coupled CLM(ED). Evaluate plant growth, forest succession, carbon fluxes in CLM(ED) and compare to field data, CLM 4.5, and ED2 Complex tropical forest structure and strata. Varying size and age classes

4 CLM(ED) Land surface divided into common-disturbance-history patches. Classified as age since disturbance in years. Patches are divided into plant functional types (PFTs) and then height classes, called cohorts. Each cohort is a group of similar plant types. Cohort based model with competition and co-existence, allowing for successional stages, size structure, competition between PFTs (Moorcroft et al. 2001). Big caveat CLM(ED) still uncalibrated, still in testing phase against data CLM Hierarchy CLM(ED) Hierarchy

5 Central Amazon field site ( ZF2 ) Observational Field data = Central Amazon, old-growth, field inventory began in ZF2 site, 100 miles north of Manaus Two 5-ha transects Model forced with meteorological tower data Comparing to single point CLM(ED) 2 nd site: Northwest Amazon, Iquitos Peru Developing pan-tropical test-bed with suite of CLM(ED) single point runs

6 Tropical Forest CLM(ED) vs. CLM4.5 vs. Observed

7 CLM(ED) vs. ED2 vs. Observed Total AGB close to observed (357 vs. 312 Mg/ha) Bias towards very high biomass in largest stem size class Aboveground Biomass Stem Density Central Amazon, CO2=367ppm AGB (Mg/ha) DBH size >80 cm

8 CLM(ED) NPP Flux NPP flux into >100 cm size class due to high stem density in >100 cm Mg C ha -1 yr -1 CLM(ED) Chambers et al Dead Wood Leaf Seed 0.9 NA Storage 0.2 NA Sapwood NA Total Realistic NPP flux pattern when weighted by stem density

9 NPP Leaf: Obs: CLM-ED: 1.58 NPP Seed: Obs: CLM-ED: 0.88 Number Density: Obs: 576 /ha CLM-ED: 172 /ha NPP: Obs: [MgC/ha/yr] CLM-ED: [MgC/ha/yr] GPP: Obs: CLM-ED: NPP AGW: Observed: CLM-ED: 4.5 NPP Fine Root: Observed: CLM-ED: 2.30

10 Mortality Rate Mortality Partitions Site: Allpahuayo-Mishana A (Peru) Basal Area Diameter Increment

11 Meteorological forcing data makes a difference Qian vs. tower forced vs. observed Tower data = less precip, less radiation, less specific humidity

12 When using different meteorological forcing data, is bias towards large trees corrected? Yes. But, diameter increment and mortality still high. CLM-ED using tower data = U shaped mortality

13 Revisit CLM(ED) NPP flux with different climate forcing Previous version (Qian) Updated (Tower data) Mg C ha -1 yr -1 CLM(ED) CLM(ED) Tower Data Chambers et al Dead Wood Leaf Seed NA Storage NA Sapwood NA Total

14 Multiple models have bias towards faster turnover rates and larger biomass increment. Why? 16 Aboveground Biomass Trend Relative to 2000; Transient CO2 (except for CLM-ED) CLM(ED)

15 What have we learned so far? How does CLM(ED) compare to CLM4.5 BGC? Lower LAI, lower AGB to more accurate values to field data, lower sensible heat flux. Forest demographic patterns in CLM(ED) need some attention (i.e., size class distribution of basal area, AGB, stem density). There is a bias towards large diameter trees. Larger trees keep getting bigger and bigger. High diameter increment and mortality rate (except for mortality when using Qian forcing data). High stem density in >100cm; leading to continual NPP into larger stems. Different meteorological forcing datasets can lead to large differences when using an ecosystem demography model. Basal Area of 18 vs. 66 m 2 ha -1 AGB of 150 vs. 600 Mg ha -1 Using site specific climate data correct bias towards large trees, but growth rates and mortality rates still high. But using site specific climate data is not feasible in global CLM(ED).

16 Next steps Pan-tropical test-bed with suite of CLM(ED) single point runs Compare against benchmarking data in Peru, Panama, Puerto Rico, Asia and more Brazilian sites Diagnosing bias towards large tropical trees, large BA, large AGB Parameter sensitivity testing of Vcmax, NPP allocation, causes of mortality, others? Investigate why high growth and mortality Global runs with competition between all PFTs Belowground nutrient constraints